Thea Presto - GPU Rendering
Thea Presto is a new engine for Thea Render, written from scratch, that uses the same characteristics and material/lighting system such as Thea itself. It is running entirely on the GPU and it is based on Nvidia CUDA.
One thing that is impressive about Thea Presto is its interactiveness that makes it fun to work with. It is ideal for rendering quick stills and animations, particularly for product design and exteriors.
With multiple GPU support and network rendering, we can render now complete sequences within a limited time frame. Thea Presto completes the list of supported engines of the most versatile renderer.
Unbiased Engine (TR1/TR2)
With the term "unbiased" we mean that rendering, seen as a simulation process, converges always to the ground truth, computing all ways of lighting transfer without any artifacts. Thea superior unbiased core is the most sophisticated in the market and delivers stunning images without any discounts.
Many render packages rely on brute force optimizations but in Thea Render we introduced innovative techniques taking advantage of algorithmic optimizations for achieving superior performance. All possible paths of lighting transfer are explored, delivering the highest accuracy.
In Thea Render, we have developed a strong mathematical framework that has led to a very robust unbiased core. This way, it can be also seen as a light simulation tool, performing calculations in the full visible spectral space and including advanced phenomena like volumetric scattering and spectral index of refraction.
One of the great advantages of using the unbiased core is that no effort is needed to setup the engine itselft. Results come out as they are in reality, without any artifacts. This way, the artist and designer can focus exclusively on bringing his ideas into life without fiddling with any engine parameters.
In addition to the above, Thea comes not with one but two, finely-tuned, engines that are optimized, in terms of performance, for different kind of scenes. Unbiased engine TR1 is optimal for exteriors and scenes with dominant direct lighting while unbiased engine TR2 is optimal for extremely difficult indirect and caustic lighting.
Having paid attention to the details, Thea unbiased core is the only one that can robustly resolve situations where other commercial unbiased renderers cannot. Such case is the sun-pool caustics problem, a particularly difficult problem to solve using unbiased methods. Another is the terminator artifact, where our solution works out-of-the-box producing smooth artifact-free renders.
Biased Engine (Adaptive BSD)
The abbreviation "BSD" comes from the word "biased" and we use this term here as an identifier of this engine that uses interpolation schemes (like irradiance cache) to render in shorter times. The word "Adaptive" on the other hand means that we implemented the engine so that more effort is put where it is needed most. Furthermore, this effort is driven by perceptual criteria in order to arrive to a high quality result according to the human perception.
Field mapping is a new proprietary technique that evaluates consistently the lighting than cannot reach easily the viewer. With traditional biased engines, the global illumination is often noisy, even when Final Gathering is coupled with Photon Mapping, resulting in blotches in the final render. With Field Mapping, evaluation becomes easy and it brings the expected high quality result without any troubles.
Due to our robust engine, we have the luxury to create presets that work in almost all cases, without further tweaking from the user. This way, the time to tune the biased engine is minimized and new users can get nice results right from the beginning. Using perceptual criteria, the engine adapts better to the local scene difficulties and the settings are simpler than before.
The issue of shadow and light leaks has been a "traditional" problem with biased render engines, especially the ones using techniques like Photon Mapping. This is for example the case, where light or shadow from one side of a wall appears on the other side as well, showing an unnatural effect. Field Mapping makes an accurate evaluation and, leaks, that are so common with Photon Mapping, are not present.
The shadow catcher enables a model to be naturally embedded inside a photograph, showing its shadow from the environmental lighting. Shadow catcher is now supported by Adaptive BSD and it is very fast! Even more, the bump and normal mapping under global illumination are also dramatically improved, giving a realistic look to bumped surfaces when indirectly lit. And this, without any speed drop.
The Adaptive BSD engine is also the best one to use when it comes to walkthrough animation. This is because the global illumination solution can be calculated for the whole sequence and reused afterwards for rendering all the frames. With our specific implementation for walkthrough animations, the engine gives a flicker-free animation without any floating "blotchy" patterns.